Table 2. Comparative results for time domain transformation-based data augmentation methods.
| Model | Permutation | Slicing | Time Warping | Window Warping | ||||
|---|---|---|---|---|---|---|---|---|
| Ave. (%) | t | Ave. (%) | t | Ave. (%) | t | Ave. (%) | t | |
| MLP | 68.17±22.17 | -3.89*** | 70.00±22.51 | -0.49 | 67.20±22.31 | -5.64*** | 69.73±22.50 | -0.87 |
| VGG | 73.75±22.00 | 0.81 | 76.33±24.47 | 2.68** | 75.50±20.82 | 1.86* | 75.88±21.63 | 2.57** |
| ResNet | 80.67±18.31 | -1.16 | 81.51±17.64 | 0.24 | 79.62±19.20 | -2.25** | 82.32±16.93 | 1.99** |
| LSTM | 49.65±26.90 | -3.56*** | 52.44±27.81 | -0.96 | 51.09±27.14 | -2.28** | 54.41±28.23 | 1.73* |
| BLSTM | 60.89±24.73 | -1.38 | 64.60±24.01 | 1.93* | 61.42±25.22 | -0.97 | 62.63±24.40 | 0.10 |
| LSTM-FCN | 81.25±17.22 | -0.74 | 80.30±19.27 | -1.67* | 78.82±18.69 | -4.66*** | 79.73±19.35 | -1.94* |
* p < 0.1,
** p < 0.05,
*** p < 0.01